Metacognition

In brief

The developmental psychologist John Flavell (born 1928) coined "metacognition" in the 1970s and set out its shape in a 1979 paper: cognition that takes cognition as its object, the mind knowing about and regulating the mind. The idea splits cleanly in two. Monitoring is the inward measurement: the feeling that an answer is on the tip of your tongue, the sense that you have studied a chapter well enough, the judgment that you have understood a proof. Control is what you do with that reading: keep searching your memory or give up, restudy the chapter or close the book, reread the proof or move on. The insight that gives the field its bite is that monitoring is not a direct pipeline to the truth about your own knowledge. It is an inference drawn from cues, and the cues can mislead. This is why fluent, easy reading can feel like learning while leaving almost nothing behind, and why the study strategies people find most comfortable are often the ones that teach them least.

The full treatment

The problem it answers

Ordinary theories of memory and reasoning describe first-order processes: how information is encoded, stored, retrieved, how inferences are drawn. They leave out the fact that the same mind also has opinions about those processes, and acts on them. A student allocating an hour of study is not just remembering; she is deciding what she does not yet know and where the hour will do the most good. A witness saying "I am certain it was him" is reporting a monitoring judgment, not a memory. Metacognition names the second-order layer that these first-order accounts omit, and asks how accurate it is and how it governs behavior.

How it works: monitoring and control

Thomas Nelson (1944 to 2005) and Louis Narens gave the field its organizing model in 1990. They pictured two interacting levels: an object level, where perceiving, remembering, and thinking happen, and a meta level, which holds a model of the object level. Two directions of flow connect them. Monitoring is information passing upward, from object to meta: the meta level is kept informed about how the object level is doing. Control is information passing downward, from meta to object: the meta level initiates, sustains, or terminates object-level activity. The framework is powerful because it separates two things everyday language runs together. You can measure how good someone's monitoring is (does their confidence track their accuracy?) independently of how they use it. Poor learning can come from a bad gauge or from a good gauge ignored.

What it claims

The central claim is that monitoring is a judgment, produced from indirect evidence, rather than a window onto the contents of memory. Asher Koriat's cue-utilization account (1997) makes this explicit: when you rate how well you have learned something, you do not consult a hidden meter reading; you infer a rating from cues such as how easily the material comes to mind, how fluently it reads, how familiar the topic feels. Some cues are diagnostic and some are not, and that is the source of both the accuracy and the errors. A related claim is directional: because control depends on monitoring, systematically biased monitoring produces systematically bad self-regulation. If ease of processing feels like mastery, learners will stop studying the moment material becomes fluent, which is often long before it becomes durable.

The key demonstrations

Two phenomena anchored the field before it had a name. The feeling of knowing was first captured experimentally by J. T. Hart in 1965: people who fail to recall an answer can still predict, better than chance, whether they will recognize it later. The judgment precedes the retrieval and is partly accurate, which shows monitoring is a real, measurable capacity, not mere talk. The tip-of-the-tongue state, studied by Roger Brown and David McNeill in 1966, is its vivid cousin. They read subjects definitions of rare words and caught people in the "TOT" state: unable to produce the word yet able to report its first letter, its number of syllables, words that sounded like it. The mind knows about a memory it cannot yet reach, monitoring running ahead of retrieval.

The practical demonstrations concern judgments of learning, the confidence ratings people give while studying. Here the news is sobering. Nate Kornell and Robert Bjork (2007) and a large body of related work show that learners' spontaneous study choices frequently backfire: people prefer massed practice and rereading, which feel productive, over spacing and self-testing, which feel harder and work better. Because the effortful methods produce less fluency in the moment, monitoring reads them as less effective, and control steers learners away from exactly what would help. Metacognition here is not a flaw in memory but a mismatch between what feels like learning and what is learning.

Related distinctions

Flavell separated metacognitive knowledge (stable beliefs, such as "I remember faces better than names") from metacognitive experiences (in-the-moment feelings, such as sudden confusion). Metamemory, the term from Nelson and Narens, is metacognition restricted to memory. And metacognition is not the same as intelligence: being able to solve problems is one thing, knowing whether you have solved them is another, and the two can come apart.

Lineage

The word is Flavell's, but the questions are older. William James, in The Principles of Psychology (1890), described the feeling of a gap in memory that is "intensely active," a state with a shape though its content is missing, which is the tip-of-the-tongue phenomenon in all but name. Jean Piaget, in whose developmental tradition Flavell worked, had studied how children come to reflect on their own reasoning. Hart's feeling-of-knowing paradigm (1965) and Brown and McNeill's tip-of-the-tongue study (1966) gave the topic a laboratory, and Flavell's synthesis in the late 1970s gave it a name and a place in developmental psychology. Nelson and Narens (1990) then supplied the monitoring-and-control architecture that most later work still uses. The practical branch, on learning strategies, grew from the memory laboratory of Robert Bjork and colleagues, whose notion of "desirable difficulties" holds that conditions which make learning feel harder often make it more durable.

The strongest case for it

The framework earns its place by being measurable and by predicting behavior. Monitoring accuracy can be quantified: line up a person's confidence judgments against their actual performance and you get calibration, a number that varies across people, tasks, and training, and that predicts how well they will regulate their own learning. The separation of monitoring from control is not a verbal trick; it dissolves real confusions, letting researchers ask whether a struggling learner has a broken gauge or a working gauge she overrides. And the applied payoff is concrete. The finding that fluency masquerades as mastery has reshaped how study skills are taught: teach learners to distrust the ease of rereading, to test themselves, to space their practice, and their outcomes improve. Few ideas in psychology connect a clean theoretical model to a demonstrable intervention as directly.

The strongest case against it

The sharpest criticism attacks the reliability of the very thing the field measures. If monitoring is an inference from cues rather than a readout, then much of what looks like insight into one's own mind is confabulation. The classic warning is Richard Nisbett and Timothy Wilson's "Telling More Than We Can Know" (1977): people readily report the workings of their own minds, and those reports are often wrong, generated after the fact from plausible theories rather than genuine access. On this view, metacognitive reports risk describing a story the mind tells about itself. The Dunning-Kruger literature makes the same point about competence: the least skilled are often the most miscalibrated, precisely because the knowledge needed to do a task is the knowledge needed to judge the doing, so the gauge fails hardest where it is needed most.

There are also boundary and construct worries. Critics note that "metacognition" has stretched to cover so much (feelings, beliefs, strategies, control decisions, confidence) that it risks losing sharp meaning, and that some measures of it correlate poorly with each other, which is what you would expect if they are tapping different things rather than one faculty. Developmental and cross-cultural findings complicate the picture too: what counts as good monitoring, and how readily it transfers across tasks, is not fixed. Finally, some argue the practical advice is oversold. Simply informing learners that testing beats rereading does not reliably change what they do, because the in-the-moment feeling of fluency is a stubborn cue that mere knowledge does not override. The gap between metacognitive knowledge and metacognitive control is itself a standing rebuke to any tidy story.

Where it stands now

Metacognition is a settled and productive area rather than a contested one, and the monitoring-and-control frame remains its backbone. The live questions have moved to the mechanism (which cues drive judgments of learning, and when they are diagnostic), to whether monitoring can be trained to transfer beyond the task it was trained on, and to how far self-regulated study can be improved by teaching people to override the pull of fluency. The applied work has spread well past the laboratory into study-skills instruction, and its central lesson is now widely repeated: the feeling of learning and the fact of learning are different quantities, and the comfortable feeling is the less trustworthy of the two. That the inner gauge is fallible is not a defeat for the field. It is the field's main finding.

Test yourself

Think of the last thing you studied by rereading it until it felt familiar, and told yourself you knew. Close the source and try to reproduce it from memory, out loud or on paper. If the familiar material turns out to be thinner than the feeling promised, you have just watched monitoring and reality come apart, which is the whole subject in a single move.